Ημερολόγιο Εκδηλώσεων

Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών

Πρόγραμμα Προπτυχιακών Σπουδών

ΠΑΡΟΥΣΙΑΣΗ ΔΙΠΛΩΜΑΤΙΚΗΣ ΕΡΓΑΣΙΑΣ

ΜΙΧΑΗΛ ΜΩΤΟΥ

με θέμα

Προσωρινή Αποθήκευση Περιεχομένου σε Κυψελωτά Δίκτυα: Μία Σύγκριση Αλγορίθμων

Content Caching in Cellular Networks: An Algorithmic Comparison

Πέμπτη 19 Σεπτεμβρίου 2019, 10 π.μ.

Αίθουσα 145.Π42, Κτίριο Επιστημών, Πολυτεχνειούπολη

Εξεταστική Επιτροπή

Καθηγητής Μιχαήλ Πατεράκης (επιβλέπων)

 Καθηγητής Αθανάσιος Λιάβας

 Αναπληρωτής Καθηγητής Αντώνιος Δεληγιαννάκης

Abstract

The ever-increasing content demand along with an increase in content size has begged the question to explore efficient caching algorithms in order to address the need for lag free content delivery and reduced costs and backhaul load. First, we consider the caching placement policy as a 0-1 Knapsack problem and then proceed to introduce the system model. Next, we comprehensively formulate the 0-1 Knapsack problem, and describe two algorithmic solutions, a simple and fast greedy algorithm, which is not optimal, and an optimal dynamic programming one. We also describe in detail the simulation model and the pertinent distributions we ran our simulations with. The results indicate that no matter the type of the weight distribution (weight here corresponds to the length of the content items) or its characteristics, the greedy algorithm manages to closely match the results of the dynamic programming one, while at the same time offering greatly reduced runtime complexity. On this basis, the simple and fast greedy algorithm considered is a very good choice as a caching policy, since content provision services currently offer large catalogues, where dynamic programming exhibits exorbitant running times.